In today’s digital world, where the content is abundant and constantly expanding across platforms, users often struggle to find content that aligns with their interests and preferences. This marks the importance of development of personalised content generation techniques that can fulfill and provide what we need by just describing it. At the same time, video content has become the most common mode for communication, marketing, and entertainment; however, its production requires resources like professional teams, equipment, and post-production editing, which makes it more of a time-consuming and expensive process. To address these challenges, this paper presents a personalized image and video synthesis framework using Conditional Generative Adversarial Networks(cGANs). This approach makes use of the CIFAR-10 data set to generate class-specific synthetic images and the UCF101 action recognition data set to generate videos. The image generation process is built on a Convolutional GAN architecture, conditioned on class labels to ensure diversity and relevance. For video generation, frames are extracted from selected video classes, processed and synthesized using a gantt trained to replicate the dynamic properties of real world actions. The proposed model achieves a competitive FID score of 50.9 demonstrating superior performance over several baseline models. Results show that cGANs are effective not only in producing visually coherent and contextually relevant images and Videos but also in enabling personalized media synthesis with minimal manual effort. They allow control Over the output by conditioning on auxiliary information, (e.g., class labels, text audio, etc.). This work lays the foundation for scalable and automated multimedia content creation, which could be applied in domains like advertising, virtual assistants, educational media and entertainment.